Modeling and Optimal Control of the Open Torque-Controlled Quadruped Robot Solo-12
| Type of publication: | Idiap-Com |
| Citation: | Adi_Idiap-Com-02-2022 |
| Number: | Idiap-Com-02-2022 |
| Year: | 2022 |
| Month: | 7 |
| Institution: | Idiap |
| Abstract: | Despite recent advances in controlling legged robots, modern quadruped robots are still not nearly as powerful as their biological counterparts. The biggest gaps exist in terms of robustness and versatility. The first is mainly caused by simplification in modeling the dynamics. On the other hand, the second point is forced by the complexity of the dynamics respectively its resulting challenging, and high computational optimization problems. Since these two requirements are in conflict with each other, a compromise must be found between the two. Depending on the task and the requirements, it is more appropriate to neglect less or more in modeling. This work aims to test and apply different controllers on the quadruped robot Solo 12. The goal is to gain knowledge about which controllers are suitable for which tasks. Additionally, it should also be reported why a controller did fail and what could be improved to achieve a better result. The robot is modeled based on centroidal dynamics or a simplified version of it to construct a whole-body controller. In addition, a method is used to plan feasible dynamical motions with the robot. Finally, there is also a Model Predictive Control (MPC) architecture tested. The robot successfully performed a trotting walk, whereby the trajectory was generated with a whole-body dynamics planner. It was also shown how large the deviation from the simulation to reality is, although complex, non-linear models were used for the dynamics. |
| Keywords: | |
| Projects |
Idiap |
| Authors | |
| Added by: | [ADM] |
| Total mark: | 0 |
|
Attachments
|
|
|
|
|
Notes
|
|
|
|
|
